NVIDIA Advances AI Infrastructure with Vera Rubin Architecture

NVIDIA is preparing for the next generation of AI infrastructure with the Vera Rubin NVL144 architecture. This new open architecture rack server is designed for gigawatt AI factories, promising increased efficiency and scalability to meet the growing demands of AI inference. Over 50 NVIDIA MGX partners are gearing up to support this innovative technology, highlighting its significance in the AI ecosystem.

Unveiling the Vera Rubin NVL144

At the OCP Global Summit, NVIDIA unveiled the specifications for the Vera Rubin NVL144 MGX-generation rack servers. These servers are engineered to support the increasing demands of AI inference, which requires robust and scalable infrastructure. The architecture features a modular design with a central printed circuit board midplane, replacing traditional cable connections to enhance assembly and serviceability.

Key Features of the Vera Rubin NVL144

Modular expansion bays for NVIDIA ConnectX-9 800GB/s networking.
NVIDIA Rubin CPX for massive-context inference.
100% liquid-cooled, energy-efficient design.

Ecosystem Support for NVIDIA Kyber

The Vera Rubin NVL144 is designed to support NVIDIA Kyber, which connects 576 Rubin Ultra GPUs. Over 20 industry partners are contributing new silicon, components, and power systems to support next-generation 800VDC data centers for the NVIDIA Kyber rack architecture. Foxconn is constructing a 40-megawatt data center in Taiwan, specifically designed for 800VDC, while other partners like CoreWeave and Oracle Cloud Infrastructure are also adopting this technology.

Vertiv's Innovative Architecture

Vertiv has introduced a space-, cost-, and energy-efficient 800 VDC MGX reference architecture, providing a complete power and cooling infrastructure solution. HPE is also announcing product support for NVIDIA Kyber and NVIDIA Spectrum-XGS Ethernet scale-across technology, further expanding the ecosystem.

Benefits of 800 VDC Infrastructure

Transitioning to 800 VDC infrastructure from traditional 415 or 480 VAC systems offers significant advantages, including increased scalability, improved energy efficiency, reduced material usage, and higher capacity for performance in data centers. This shift is crucial for supporting the growing demands of AI workloads.

Scaling AI Factories with Vera Rubin NVL144

The Vera Rubin NVL144 MGX compute tray features an energy-efficient, modular design that supports the scaling of AI factories. NVIDIA plans to contribute these innovations as an open standard for the OCP consortium, allowing partners to mix and match modular components and accelerate scaling with the architecture.

The Future of NVIDIA Kyber

The OCP ecosystem is preparing for NVIDIA Kyber, the successor to NVIDIA Oberon, which will house 576 NVIDIA Rubin Ultra GPUs by 2027. NVIDIA Kyber is engineered to boost rack GPU density, scale up network size, and maximize performance for large-scale AI infrastructure. It will become a foundational element of hyperscale AI data centers, enabling superior performance, efficiency, and reliability for generative AI workloads.

NVIDIA NVLink Fusion Ecosystem

NVIDIA NVLink Fusion is gaining traction, enabling companies to integrate their semi-custom silicon into optimized data center architectures. Intel and Samsung Foundry are joining the NVLink Fusion ecosystem, allowing AI factories to scale up quickly for demanding workloads. With over 20 partners contributing to this ecosystem, the future of gigawatt AI factories looks promising.